Kristen Berman
CEO and co-founder of Irrational Labs, co-founder of Common Cents Lab
Where to hear and read about Kristen Berman1 episode or test
Kristen Berman is CEO and co-founder of Irrational Labs, a behavioral science consultancy she started in 2013 with Dan Ariely, and co-founder of Common Cents Lab. The thing most worth learning from her is to define one uncomfortably specific behavior, then remove its barriers and give an immediate benefit for doing it.
In order to change behavior you have to pick a behavior that you want to change.
Using behavioral science to improve your product | Kristen Berman (Irrational Labs)
Kristen Berman: the story and the lessons
1Kristen Berman's path from QuickBooks PM to Irrational Labs
Berman started as a product manager guessing at behavior from a few interviews. She was a PM on QuickBooks Online at Intuit, drawing insight from five to ten customer conversations.
She was struck that the product community was missing decades of papers and studies on how people behave.
She met Dan Ariely around 2008, just after Predictably Irrational came out. That meeting led her into behavioral economics and later into building a team with him.
Three years embedded at Google gave her early proof. She and Ariely started Google's behavioral economics group, hired internal Googlers and worked with over 25 teams, from self-driving cars to YouTube. Credit Karma recurring deposit setup rose 18%.
Google now has one of the bigger behavioral science teams.
Irrational Labs grew from a handful of clients to 20 behavioral scientists. She started it in 2013 with Ariely. It designs and tests changes to user behavior, with clients such as Microsoft, LinkedIn and many startups, across industries and not just tech.
She also taught the process to others through an eight-week Behavioral Design Bootcamp with 16 modules, Slack feedback and office hours.
2Berman's 3B framework: behavior, barriers, benefits
Berman boils hundreds of biases down to three questions. Which behavior, what barriers, what benefits. Her team calls biases and heuristics psychologies, and the model groups the ones most useful to PMs and marketers.
The model comes from repeated use with hundreds of companies. Google, Microsoft and LinkedIn use it.
Berman defines the target behavior uncomfortably specifically. Her Peloton example is two 10-minute workouts with two different instructors within seven days of starting the app. Without that definition a team cannot name the psychologies shaping the decision.
She says the only wrong answer is "log in", and that teams should argue about the definition.
She splits barriers into logistical and cognitive. Logistical barriers are things like form fields. Cognitive ones are uncertainty, status quo and information aversion. At Lyft, not knowing when the car arrives is the cognitive part, so riders open Uber.
She avoids saying always about human behavior because it is very complex, so she says more often than not.
Berman gives every step an immediate benefit, because people are present biased. At Asana, logging a task pays off later in an on-time project. The checkbox (completion bias) and the teammate notification (social desirability) pay off now.
The unicorn confetti is secondary. She says the sense of being done is the powerful part. The behavior usually matches the activation milestone, worked back from what correlates with retention.
3Why Berman does not trust requested features
A heavily requested budgeting feature moved spending by zero. Support forums and interviews all asked for budgeting at a fintech app. Berman's team got them to run a 10,000-person test with a control that only showed spend and two budgeting designs. Average spend and spend variability did not change.
Budgeting needs a budget, a plan and repeated action all month, so it rarely changes behavior. She says to be skeptical of any work put on the user.
She says behavior follows what people do, not what they intend. Attitudes like wanting to be a runner do not drive behavior. Changing behavior takes a redesigned environment, not a stronger goal.
She calls the self-help message that a goal is enough the big annoying thing about that industry.
Her first line of defense is a default or rule of thumb. Auto-enrollment in the 401K is why Americans have retirement savings. Defaults need back-end work, so a rule of thumb, such as no lifts on weekdays and lifts on weekends, is the fallback.
A rule of thumb replaces weighing "is three dollars worth it" every time. It is imperfect, but people stick with it.
4How Berman runs a behavioral design project
She starts with a short workshop to agree on one behavior. Each teammate defines engagement separately and the team compares answers. They usually differ, and a team that cannot agree cannot build a product that tells users what to do.
Once aligned, the psychology behind the behavior becomes an easier conversation.
Berman runs a behavioral diagnosis, a journey map on steroids. The team screenshots every step to the behavior and overlays the psychology at each one. For an existing product the output is a deck of 200 or 300 slides, and the flow may have 50 steps.
It studies what people do, not what they say they will do. In a health app, a step might be information aversion, where users avoid seeing their test result. Teams often have light bulbs just from doing it.
She reads the research before talking to users. For a new problem she advises a day on Google Scholar before talking to five users, searching the field's buzzwords such as chronic care to find a meta-analysis.
The TikTok team began this way and found that people share less when reminded of accuracy at the moment of sharing.
She tests versions against each other, never one design alone. Users can like or hate every design, so a single option tells you nothing. The test measures which version drives the behavior, not whether people like it.
The TikTok pre-test de-risked the launch.
5Berman's conversion tactics: questions, deadlines and resuming
Berman adds an easy question to sign-up flows to lift conversion. A question adds friction but also motivation. Trunk Club reported a 133% conversion increase. On TytoCare, 53% of quiz completers bought versus 37% of non-completers.
The question can replace a benefit carousel and need not be one you need answered. Removing an open text field improved conversion 40%, since dropdowns and multiple choice are easier.
She makes the question a real setting that shows the benefit. A bill pay app should ask whether to pay bills immediately, weekly or monthly. That sets up the app and implies immediate payment is possible.
Apartment List does the same, asking about studios, patios and views.
She treats deadlines as a gift. Deadlines help people prioritize things they already want to do. Adding one raised borrower sign-up conversion at Kiva Zip.
A nonprofit refused a deadline for interview signups as too hurtful, but across multiple companies deadlines increased completion.
She returns dropouts to the step where they left. For long sign-up flows, reminders should land people where they abandoned, not at the beginning.
Wealthfront found that restarting people at the beginning was frustrating.
She uses right-for-wrong motivators, but only toward good outcomes. Vaccines were pushed with donuts, free marijuana and Ohio lottery tickets, and Peloton uses streaks and instructor shout-outs. Clubhouse's two status tiers pushed people to follow more people.
She says to check that the end result benefits the customer, and to use error-message tactics sparingly because users will not love an app full of them.
6How Berman lifted One Medical onboarding bookings by 20%
She defined engagement as booking a first appointment during signup. A new member who gets sick six months later will call their old doctor, so the app must set its purpose at signup.
The barriers were choosing a provider, finding a time and not knowing what One Medical covers. The benefit is getting health questions answered immediately.
Berman cut choices to one recommended provider and a few slots. Onboarding asked a few questions about mental, sleep and physical health, recommended one provider with a reason, offered next-day times and suggested a virtual visit. Bookings rose 20%.
7How Berman cut misinformation shares on TikTok by 24%
She added friction at the moment of sharing. The team chose shares as the behavior, not likes or comments. They added an unverified label and an "are you sure" pop-up on share, and misinformation shares fell 24%.
The reminder worked at the point of sharing, not before or after. She compares it to an elevator study where doors closing 16 seconds slower pushed people to the stairs. Facebook and Twitter use similar tactics.
The team narrowed about 30 ideas to two live conditions. They tested five pop-up versions on over 1,000 users from Prolific, then launched two conditions plus a control in the product.
The feature is still live. It was one of the first published studies showing these tactics reduce misinformation.
8Berman on incentives, ethics and hiring
Berman was paid for a conversion bump and found herself suggesting predatory ideas. Lending Club offered a large payout for a five-point bump in borrower conversion, and nothing otherwise. She was also leading Duke's Common Cents Lab, which builds savings apps for low- to moderate-income Americans.
Legal blocked the predatory ideas. Her team got 4.5 of the five points and missed the payout. She says "we are what we measure".
She sets incentives on the customer-aligned behavior over a long horizon. Behavior aligns with customer outcomes better than active use or retention does, which she keeps tracking but does not make the incentive.
Quarterly targets push short-term bumps, so she advises annual horizons.
Berman does not trust interviews to predict job performance. She says there is little evidence they do, so she uses skills assessments and trials, and uses interviews for culture fit.
Her favorite question asks for a personality trait that is both a strength and a weakness. Her own is confidence, which can mean not listening.
Using behavioral science to improve your product | Kristen Berman (Irrational Labs)
Kristen Berman explains the 3B framework (behavior, barriers, benefits) and shows how behavioral diagnosis, immediate benefits and added friction change user behavior. Her team cut misinformation shares on TikTok by 24%, raised One Medical onboarding bookings by 20%, and showed a heavily requested budgeting feature moved spending by zero.
Berman started as a product manager guessing at behavior from a few interviews. She was a PM on QuickBooks Online at Intuit, drawing insight from five to ten customer conversations.
She was struck that the product community was missing decades of papers and studies on how people behave.
She met Dan Ariely around 2008, just after Predictably Irrational came out. That meeting led her into behavioral economics and later into building a team with him.
Three years embedded at Google gave her early proof. She and Ariely started Google's behavioral economics group, hired internal Googlers and worked with over 25 teams, from self-driving cars to YouTube. Credit Karma recurring deposit setup rose 18%.
Google now has one of the bigger behavioral science teams.
Irrational Labs grew from a handful of clients to 20 behavioral scientists. She started it in 2013 with Ariely. It designs and tests changes to user behavior, with clients such as Microsoft, LinkedIn and many startups, across industries and not just tech.
She also taught the process to others through an eight-week Behavioral Design Bootcamp with 16 modules, Slack feedback and office hours.
Berman boils hundreds of biases down to three questions. Which behavior, what barriers, what benefits. Her team calls biases and heuristics psychologies, and the model groups the ones most useful to PMs and marketers.
The model comes from repeated use with hundreds of companies. Google, Microsoft and LinkedIn use it.
Berman defines the target behavior uncomfortably specifically. Her Peloton example is two 10-minute workouts with two different instructors within seven days of starting the app. Without that definition a team cannot name the psychologies shaping the decision.
She says the only wrong answer is "log in", and that teams should argue about the definition.
She splits barriers into logistical and cognitive. Logistical barriers are things like form fields. Cognitive ones are uncertainty, status quo and information aversion. At Lyft, not knowing when the car arrives is the cognitive part, so riders open Uber.
She avoids saying always about human behavior because it is very complex, so she says more often than not.
Berman gives every step an immediate benefit, because people are present biased. At Asana, logging a task pays off later in an on-time project. The checkbox (completion bias) and the teammate notification (social desirability) pay off now.
The unicorn confetti is secondary. She says the sense of being done is the powerful part. The behavior usually matches the activation milestone, worked back from what correlates with retention.
A heavily requested budgeting feature moved spending by zero. Support forums and interviews all asked for budgeting at a fintech app. Berman's team got them to run a 10,000-person test with a control that only showed spend and two budgeting designs. Average spend and spend variability did not change.
Budgeting needs a budget, a plan and repeated action all month, so it rarely changes behavior. She says to be skeptical of any work put on the user.
She says behavior follows what people do, not what they intend. Attitudes like wanting to be a runner do not drive behavior. Changing behavior takes a redesigned environment, not a stronger goal.
She calls the self-help message that a goal is enough the big annoying thing about that industry.
Her first line of defense is a default or rule of thumb. Auto-enrollment in the 401K is why Americans have retirement savings. Defaults need back-end work, so a rule of thumb, such as no lifts on weekdays and lifts on weekends, is the fallback.
A rule of thumb replaces weighing "is three dollars worth it" every time. It is imperfect, but people stick with it.
She starts with a short workshop to agree on one behavior. Each teammate defines engagement separately and the team compares answers. They usually differ, and a team that cannot agree cannot build a product that tells users what to do.
Once aligned, the psychology behind the behavior becomes an easier conversation.
Berman runs a behavioral diagnosis, a journey map on steroids. The team screenshots every step to the behavior and overlays the psychology at each one. For an existing product the output is a deck of 200 or 300 slides, and the flow may have 50 steps.
It studies what people do, not what they say they will do. In a health app, a step might be information aversion, where users avoid seeing their test result. Teams often have light bulbs just from doing it.
She reads the research before talking to users. For a new problem she advises a day on Google Scholar before talking to five users, searching the field's buzzwords such as chronic care to find a meta-analysis.
The TikTok team began this way and found that people share less when reminded of accuracy at the moment of sharing.
She tests versions against each other, never one design alone. Users can like or hate every design, so a single option tells you nothing. The test measures which version drives the behavior, not whether people like it.
The TikTok pre-test de-risked the launch.
Berman adds an easy question to sign-up flows to lift conversion. A question adds friction but also motivation. Trunk Club reported a 133% conversion increase. On TytoCare, 53% of quiz completers bought versus 37% of non-completers.
The question can replace a benefit carousel and need not be one you need answered. Removing an open text field improved conversion 40%, since dropdowns and multiple choice are easier.
She makes the question a real setting that shows the benefit. A bill pay app should ask whether to pay bills immediately, weekly or monthly. That sets up the app and implies immediate payment is possible.
Apartment List does the same, asking about studios, patios and views.
She treats deadlines as a gift. Deadlines help people prioritize things they already want to do. Adding one raised borrower sign-up conversion at Kiva Zip.
A nonprofit refused a deadline for interview signups as too hurtful, but across multiple companies deadlines increased completion.
She returns dropouts to the step where they left. For long sign-up flows, reminders should land people where they abandoned, not at the beginning.
Wealthfront found that restarting people at the beginning was frustrating.
She uses right-for-wrong motivators, but only toward good outcomes. Vaccines were pushed with donuts, free marijuana and Ohio lottery tickets, and Peloton uses streaks and instructor shout-outs. Clubhouse's two status tiers pushed people to follow more people.
She says to check that the end result benefits the customer, and to use error-message tactics sparingly because users will not love an app full of them.
She defined engagement as booking a first appointment during signup. A new member who gets sick six months later will call their old doctor, so the app must set its purpose at signup.
The barriers were choosing a provider, finding a time and not knowing what One Medical covers. The benefit is getting health questions answered immediately.
Berman cut choices to one recommended provider and a few slots. Onboarding asked a few questions about mental, sleep and physical health, recommended one provider with a reason, offered next-day times and suggested a virtual visit. Bookings rose 20%.
She added friction at the moment of sharing. The team chose shares as the behavior, not likes or comments. They added an unverified label and an "are you sure" pop-up on share, and misinformation shares fell 24%.
The reminder worked at the point of sharing, not before or after. She compares it to an elevator study where doors closing 16 seconds slower pushed people to the stairs. Facebook and Twitter use similar tactics.
The team narrowed about 30 ideas to two live conditions. They tested five pop-up versions on over 1,000 users from Prolific, then launched two conditions plus a control in the product.
The feature is still live. It was one of the first published studies showing these tactics reduce misinformation.
Berman was paid for a conversion bump and found herself suggesting predatory ideas. Lending Club offered a large payout for a five-point bump in borrower conversion, and nothing otherwise. She was also leading Duke's Common Cents Lab, which builds savings apps for low- to moderate-income Americans.
Legal blocked the predatory ideas. Her team got 4.5 of the five points and missed the payout. She says "we are what we measure".
She sets incentives on the customer-aligned behavior over a long horizon. Behavior aligns with customer outcomes better than active use or retention does, which she keeps tracking but does not make the incentive.
Quarterly targets push short-term bumps, so she advises annual horizons.
Berman does not trust interviews to predict job performance. She says there is little evidence they do, so she uses skills assessments and trials, and uses interviews for culture fit.
Her favorite question asks for a personality trait that is both a strength and a weakness. Her own is confidence, which can mean not listening.